【Python】OpenCV库学习笔记（四）

r囧r小猫 2023-01-07 12:29 104阅读 0赞

### OpenCV库学习笔记（四） ###

*  1. ROI
 *  2. 泛洪填充
 *  3. 模糊操作
 *   *  3.1 均值模糊
     *  3.2 中值模糊
     *  3.2 自定义模糊
 *  4. 高斯模糊
 *  5. 边缘保留滤波
 *   *  5.1 高斯双边
     *  5.2 均值迁移

# 1. ROI #

*  ROI(**region of interest**)，也就是感兴趣区域，如果你设置了图像了ROI，那么在使用OpenCV的函数的时候，会只对ROI区域操作，其他区域忽略。

import cv2 as cv
    import numpy
    
    image = cv.resize(cv.imread("star.jpg"), (400, 300))
    cv.imshow("image", image)
    # 选取出目标区域
    region = image[50:200, 100:250]
    # 将目标区域变成灰色
    gray = cv.cvtColor(region, cv.COLOR_BGR2GRAY)
    # 再将目标区域放回原图
    image[50:200, 100:250] = cv.cvtColor(gray, cv.COLOR_GRAY2BGR)
    cv.imshow("image", image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70]

# 2. 泛洪填充 #

def floodFill(image, mask, seedPoint, newVal, loDiff=None, upDiff=None, flags=None)

*  image表示原图
 *  mask表示遮罩图
 *  seedPoint表示填充起始坐标
 *  newVal表示需要填充的颜色
 *  loDiff表示需要填充颜色的像素值的下限
 *  upDiff表示需要填充颜色的像素值的上限
 *  flags表示填充方法，彩色图像填充必须为FLOODFILL\_FIXED\_RANGE，二值图像使用

> 假设seedPoint的值为（R，G，B），遍历整张图片，将所有像素点的值位于（R-loDiff，G-loDiff，B-loDiff）~（R+upDiff，GR+upDiff，BR+upDiff）的点全部变成newVal颜色

**（1）填充彩色图像**

import cv2 as cv
    import numpy
    
    
    def flood_fill_demo(image):
        copy_image = image.copy()   # 拷贝图片
        h, w = image.shape[:2]    # 获取图片宽和高
        mask = numpy.zeros([h + 2, w + 2], numpy.uint8)  # 获取一张mask，其大小必须比原图长和宽大2
        cv.floodFill(copy_image, mask, (30, 30), (0, 255, 255), (100, 100, 100), (50, 50, 50), cv.FLOODFILL_FIXED_RANGE)
        cv.imshow("flood_fill_image", copy_image)
    
    image = cv.resize(cv.imread("star.jpg"), (400, 300))
    cv.imshow("image", image)
    flood_fill_demo(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 1]

（2）填充二值图像

import cv2 as cv
    import numpy
    
    def fill_binary_demo():
        # 创建一张中间白周围黑的图片
        image = numpy.zeros([400, 400, 3], numpy.uint8)
        image[100:300, 100:300, :] = 255
        cv.imshow("image", image)
    
        # 创建遮罩图
        mask = numpy.ones([402, 402, 1], numpy.uint8)
        mask[101:301, 101:301] = 0
        # 从（200,200）点向周围填充(0, 0, 255)这个颜色
        cv.floodFill(image, mask, (200, 200), (0, 0, 255), cv.FLOODFILL_MASK_ONLY)
        cv.imshow("filled image", image)
    
    fill_binary_demo()
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 2]

# 3. 模糊操作 #

> 模糊操作原理：
> 
>  *  基于离散卷积
>  *  定义好每个卷积核
>  *  不同卷积核得到不同的卷积效果
>  *  模糊是卷积的一种表象

卷积计算原理：

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 3]

## 3.1 均值模糊 ##

*  可以用于去除

import cv2 as cv
    import numpy
    
    
    def blur_demo(image):
        dst = cv.blur(image, (20, 1))
        cv.imshow("dst", dst)
    
    image = cv.imread("01.jpg")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    blur_demo(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 4]

## 3.2 中值模糊 ##

*  可以用于去除`椒盐噪声`

> 原图中有许许多多的小黑点，使用中值模糊可以将小黑点给去除，这就叫`去除椒盐噪声`

import cv2 as cv
    import numpy
    
    def mid_blur(image):
        dst = cv.medianBlur(image, 5)
        cv.imshow("dst", dst)
    
    image = cv.imread("test1.png")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    mid_blur(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 5]

## 3.2 自定义模糊 ##

*  主要自定义卷积核

import cv2 as cv
    import numpy
    
    def cus_blur(image):
        # 自定义初始化卷积核
        kernel = numpy.ones([17, 17], numpy.float32) / 200
        dst = cv.filter2D(image, -1, kernel=kernel)
        cv.imshow("dst", dst)
    
    image = cv.imread("test1.png")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    cus_blur(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 6]

*  还可以用于锐化（增强立体感）

import cv2 as cv
    import numpy
    
    def cus_blur(image):
        kernel = numpy.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]], numpy.float32)
        dst = cv.filter2D(image, -1, kernel=kernel)
        cv.imshow("dst", dst)
    
    image = cv.imread("01.jpg")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    cus_blur(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 7]

# 4. 高斯模糊 #

> 二维高斯分布

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 8]

> 立体高斯分布

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 9]

*  代码层面

import cv2 as cv
    import numpy
    
    
    def clamp(pv):
        if pv > 255:
            return 255
        if pv < 0:
            return 0
        return pv
    
    def gaussion_noise(image):
        h, w, c = image.shape
        for row in range(w):
            for col in range(h):
                s = numpy.random.normal(0, 20, 3)
                B = image[row, col, 0]
                G = image[row, col, 1]
                R = image[row, col, 2]
                image[row, col, 0] = clamp(B + s[0])
                image[row, col, 1] = clamp(G + s[1])
                image[row, col, 2] = clamp(R + s[2])
        cv.imshow("noise image", image)
    
    image = cv.imread("01.jpg")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    gaussion_noise(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 10]

# 5. 边缘保留滤波 #

## 5.1 高斯双边 ##

def bilateralFilter(src, d, sigmaColor, sigmaSpace, dst=None, borderType=None)

*  src表示图像
 *  d表示distance，可以设置为0
 *  sigmaColor取大一点，为了将小的噪声给模糊掉
 *  sigmaSpace取小一点

import cv2 as cv
    import numpy
    
    
    def demo(image):
        dst = cv.bilateralFilter(image, 0, 100, 15)
        cv.imshow("dst",dst)
    
    image = cv.imread("01.jpg")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    demo(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 11]

## 5.2 均值迁移 ##

def pyrMeanShiftFiltering(src, sp, sr, dst=None, maxLevel=None, termcrit=None)

import cv2 as cv
    import numpy
    
    def demo(image):
        dst = cv.pyrMeanShiftFiltering(image, 10, 50)
        cv.imshow("dst",dst)
    
    image = cv.imread("star.jpg")
    cv.namedWindow("input image", cv.WINDOW_AUTOSIZE)
    cv.imshow("input image", image)
    demo(image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 12]

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70]: /images/20221119/7ed3beda854f4d7cb6652c1a14bcc09c.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/20210117231404534.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20210117231426486.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 3]: https://img-blog.csdnimg.cn/20210117231441505.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 4]: https://img-blog.csdnimg.cn/20210117231508255.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 5]: https://img-blog.csdnimg.cn/20210117231551733.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 6]: https://img-blog.csdnimg.cn/20210117231608905.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 7]: https://img-blog.csdnimg.cn/20210117231635623.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 8]: https://img-blog.csdnimg.cn/20210117231654299.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 9]: https://img-blog.csdnimg.cn/20210117231710343.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 10]: https://img-blog.csdnimg.cn/20210117231730305.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 11]: https://img-blog.csdnimg.cn/20210117231812536.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg_size_16_color_FFFFFF_t_70 12]: https://img-blog.csdnimg.cn/2021011723183042.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NTQzNzAyMg==,size_16,color_FFFFFF,t_70